Fig 1.
Sample distributions of motorist households’ fuel purchases and mileage, NTS 2002–2008.
Notes: 1. The NTS reports mileage separately for each vehicle. The figure is obtained by aggregating over vehicles. 2. Censored at the 99th percentile.
Table 1.
Extent of infrequency of purchase by income quintile.
Fig 2.
Coefficient plot of probit regression models of fuel purchase in the diary week.
Notes. 1. Probit coefficients are shown as diamonds with lines representing confidence intervals. 2. Continuous regressors are standardised to have mean zero and unit variance. 3. Household-level variables are derived from individual- and vehicle-level data (authors’ calculations). 4. A dummy for each top-coded variable is included in the estimations but not shown. 5. For model 1 McFadden’s pseudo r-sq (adjusted) = 0.053, count r-sq (adjusted) = 0.11 and Log-L = -27250.28. Model 2 has 2 additional parameters; McFadden’s pseudo r-sq (adjusted) = 0.078, count r-sq (adjusted) = 0.16. and Log-L -26534.82. The LLR statistic is therefore 1430.92~χ2(2); p<0.001.
Fig 3.
Estimated propensity scores: kernel density estimates.
Note: Epanechnikov kernel.
Fig 4.
Standardised percentage bias between covariates in Z = 0 and Z = 1 households.
Notes. 1. Covariates are those shown in Fig 2, less square and interaction terms, plus topcode dummies. 2. The mean, median and maximum absolute standardised % bias are 1.1, 0.6 and 22.7 for model 1 and 1.0, 0.7 and 3.3 for model 2.
Fig 5.
Quantities at the pump (litres) derived from PSM using model 2.
Fig 6.
Preferred estimates of fuel consumption rates, derived from model 2.
Notes 1. Kernel density estimates (Epanechnikov kernel) overlaid. 2. Excludes top percentile.
Table 2.
Estimates from PSM based imputation, recorded fuel purchase and mileage.
Fig 7.
Quantile-quantile plots of estimated fuel consumption rates against recorded annual mileage.
Note Values have been divided by the maximum of the range.
Table 3.
Decile shares of UK motorist households’ CO2 emissions from motor fuels.
Fig 8.
Estimated monetary effects of a carbon tax or ration at £100/tCO2.
Note: ‘diary sample’ weights applied, recalculated for the pooled sample.
Fig 9.
Estimated monetary effects of a ‘cap and share’ or ‘tax and dividend’ policy at £100/tCO2.
Note: ‘diary sample’ weights applied, recalculated for the pooled sample.